(Virtual) Seminar Series | The Role of Subset Selection in Matching Markets

Seminar by Alfredo Torrico, Postdoctoral Researcher

CERC in Data Science, Polytechnique Montreal

The study of matching markets has become increasingly important due to its wide range of applications such as online marketplaces, school admission and kidney exchange. These intricate socio-technical systems require novel approaches that can deal with complex technical and societal challenges. In this talk, I will focus on matching market models that include a subset selection component in their design as an algorithmic alternative to improve the quality of the outcome. The first, and main, part of the talk will be concentrated on the use of assortment optimization as a tool to improve the efficiency of a decentralized two-sided matching market model. I will discuss efficient algorithms with provable guarantees that construct assortments for one side of the market by accounting for the structure of the preferences of both sides. The second part of the talk will be focused on the school admission matching market and the impact that the capacity expansion of a subset of schools has in the quality of the students’ allocation. In this section, I will briefly discuss the complexity of the problem and the algorithmic approaches proposed in our work.

This talk is based on joint works with F. Bobbio, M. Carvalho and A. Lodi.

Alfredo Torrico is a postdoctoral researcher at the CERC in Data Science, Polytechnique Montreal, under the supervision of Margarida Carvalho and Andrea Lodi. In 2019, he obtained his Ph.D. in Operations Research in the School of Industrial and Systems Engineering at Georgia Tech, advised by Mohit Singh, Sebastian Pokutta and Alejandro Toriello. His main research interests lie at the intersection of discrete optimization and operations research, with a focus on matching markets and subset selection problems. He is also interested in (1) societal challenges that arise in emerging platforms such as fairness, diversity, and the spread of misinformation; (2) the interactions between discrete optimization and machine learning such as the design of new approaches for submodular optimization.